Professional Documents
Culture Documents
Processes 05 00034 PDF
Processes 05 00034 PDF
Processes 05 00034 PDF
Review
Review of Field Development Optimization of
Waterflooding, EOR, and Well Placement Focusing on
History Matching and Optimization Algorithms
Jackson Udy, Brigham Hansen, Sage Maddux, Donald Petersen, Spencer Heilner, Kevin Stevens,
David Lignell and John D. Hedengren *
Department of Chemical Engineering, Ira A. Fulton College of Engineering and Technology,
Brigham Young University, 350 Clyde Building, Provo, UT 84602, USA; jacksonudy@byu.edu (J.U.);
brighamhansen@byu.edu (B.H.); sage.maddux@irsri.org (S.M.); donaldep@byu.edu (D.P.);
spencerjh1@gmail.com (S.H.); kevin.stevens@irsri.org (K.S.); davidlignell@byu.edu (D.L.)
* Correspondence: john.hedengren@byu.edu; Tel.: +1-801-477-7341
Abstract: This paper presents a review of history matching and oil field development optimization
techniques with a focus on optimization algorithms. History matching algorithms are reviewed as a
precursor to production optimization algorithms. Techniques for history matching and production
optimization are reviewed including global and local methods. Well placement, well control, and
combined well placement-control optimization using both secondary and tertiary oil production
techniques are considered. Secondary and tertiary recovery techniques are commonly referred to as
waterflooding and enhanced oil recovery (EOR), respectively. Benchmark models for comparison of
methods are summarized while other applications of methods are discussed throughout. No single
optimization method is found to be universally superior. Key areas of future work are combining
optimization methods and integrating multiple optimization processes. Current challenges and
future research opportunities for improved model validation and large scale optimization algorithms
are also discussed.
1. Introduction
Researchers and practitioners are investigating methods to increase production yield from existing
reservoirs. Currently, the majority of reservoirs economically produce at most 60% of the original
in place volumes throughout the life of the reservoir using enhanced oil recovery techniques [1].
With optimized waterflooding and enhanced oil recovery methods, there is potential to increase the
amount of petroleum that can be economically produced from these reserves.
Enhanced Oil Recovery (EOR) is the practice of using various techniques to increase the amount of
oil and gas that can be extracted from existing resources. Many different techniques have been developed
for this purpose such as steam flooding [2–6], polymer flooding [2,7–13], and gas injection [2,14,15],
among others. In this paper, EOR is defined as the injection of gas, steam, surfactants, or other
chemicals into the reservoir to improve reservoir productivity. EOR is considered a tertiary production
technique usually following primary and secondary production phases. In primary production
natural forces within the reservoir (fluid expansion, rock expansion, aquifer influx, etc.) provide
adequate energy to allow for the extraction of oil and gas. Secondary production is also called
waterflooding. In this phase of production water is injected into the reservoir to increase production.
Different waterflooding and EOR strategies may be chosen based on reservoir properties and the
current production stage of the reservoir. Often these strategies are chosen in order to maximize
total oil production or net present value (NPV). Reservoir and production engineers often use trial
and error via computer simulations to determine the allocation of production resources. However,
this method reduces the probability of finding the optimal field development and production scheme
as relatively few scenarios can be considered. This inability to run large numbers of simulations leads
to suboptimal reservoir planning and reduced production. In addition to computational limitations,
other significant challenges are frequently faced when trying to optimize reservoir development.
Some of these include lack of adequate historical data to calibrate reservoir models, large uncertainty
in physical properties, and unpredictable oil and gas prices among others. Finding the optimal field
development and production scheme increases the productivity of the reservoir. To achieve this,
in light of aforementioned challenges, it is necessary to make use of computer-aided optimization
techniques to more fully assess reservoirs and make improved field development and production
plans. The algorithms for typical field development numerical optimization are shown in Figure 1.
The reservoir system (top box) is composed of several injectors and producers. Adjusting the system
inputs (i.e., number of new production wells, number new of injection wells, location of new wells,
injection flow rate, injection composition, production flow rate, etc.) over the operating life of the
reservoir leads to changes in the instantaneous outputs (i.e., flow rate, total recovered oil, produced
water, etc.). Noise due to measurement or actuator inaccuracy is introduced into the process as
inputs or outputs to the reservoir thereby increasing the uncertainty. Sensor data is fed to history
matching algorithms that seek to adjust parameters of the simulated reservoir system to match the
actual reservoir. This history matching process is the first optimization method discussed in this review
paper. Once a sufficiently accurate representation of the reservoir is achieved, the system model or a
simplified version of the model is optimized and new field development and operation strategies are
calculated. These strategies are then implemented with the actual reservoir for maximized economics
or other objectives related to waterflooding or EOR [16]. Throughout this paper optimization dealing
with the variation of well operation parameters for either production or injection wells (i.e., production
rate, injection rate, injection composition, etc.) is referred to as well control optimization.
Reservoir
Noise Input Output Noise
System
New Input
Optimization
System Model Sensors
Algorithm
History
Matching
The research on field development optimization is extensive. It is impractical to review every facet
of production optimization in a single review paper. Therefore, this review focuses on two key areas
related to production optimization: history matching and optimization algorithms. Specifically, these
Processes 2017, 5, 34 3 of 25
key areas are explored as they pertain to the optimization of waterflood, EOR, and well placement.
A contribution of this work is the combined review of history matching and production optimization
as tightly linked in practice but rarely reviewed as a holistic process of increasing NPV. This review
examines the state of the art in production optimization algorithms and finds that there is no one
optimization method in use that is universally superior to others. Combined well placement and well
control optimization techniques are also thoroughly reviewed. The current benchmark models for
applications of waterflood and EOR optimization are also discussed. Future work recommendations
in optimization methods and applications are made. Where a specific topic is too broad to cover in
this review, related review papers [17–24] provide further detail for the interested reader. In particular,
one area of active research related to production optimization is the quantification of uncertainty, this is
reviewed by Oliver and Chen [22]. Future reviews could include an assessment of recent work in
this area.
History matching is a data reconciliation process where reservoir data (production, injection,
geological, etc.) is used to to tune uncertain reservoir model parameters. The resulting reservoir
model is then used in field development and production optimization to determine how a reservoir
should be developed and operated to maximize relevant production objectives, such as project
profitability. Various optimization algorithms are used in both the history matching data reconciliation
process (finding optimum values of uncertain parameters) and the production optimization process
(determining optimum development and operation strategies). A general mathematical expression of
both data reconciliation and production optimization is given by the spatially discretized Equation (1)
and is developed further in subsequent sections.
δx
0 = f( , y, x, p, u) (1b)
δt
0 = g(y, x, p, u) (1c)
0 ≤ h(y, x, p, u) (1d)
In this case, y ∈ Rs are the measured or optimized states (outputs); x ∈ Rq are the state variables;
p ∈ <q are a set of unknown or uncertain parameters; and u ∈ Rm are a set of inputs (i.e., injector flow
rates, compositions, locations, etc.). Reservoir dynamic expressions ( f ), equations that relate reservoir
states to measurements ( g), and inequality constraints (h) compose the general set of equations that
are solved in history matching or production optimization. The scalar objective or multi-objective
function (Ψ) takes a different form whether the goal is to match measured values or maximize the
economics of the reservoir. Continuous (i.e., valve position between 0–100%) or discrete (i.e., pump on
or off) variables are either input into the model or else determined by the optimizer. Algebraic, ordinary
differential, and partial differential equations are permitted in the general form of Equation (1) [25–29].
Section 2 reviews history matching techniques as applied to enhanced oil recovery and
waterflooding. Section 3 surveys current state-of-the-art optimization techniques that are under
development or applied in practice. Various methods and their applications are considered in each
section. The concluding remarks cover the current state of the field as well as future areas of research.
Equation (2). Here no is the norm for the minimization (i.e., no = 1 is the 1-norm), α is a weighting on
measurement error, and β is a penalty on deviation from prior parameter values. Note that the form of
the model Equations (2b)–(2d) depends on choice of physics-based, empirical, or reduced order model
form [30].
Ψ = min α
y p − ym
n + β k p − p0 kno
(2a)
p,y p ,x o
δx
s.t. 0 = f ( , y, x, p, u) (2b)
δt
0 = g(y, x, p, u) (2c)
0 ≤ h(y, x, p, u) (2d)
Once this difference between predicted (y p ) and measured (ym ) values is minimized, it is assumed
that the model can be used to predict the future performance of the reservoir. The main difficulty with
history matching is that it is an inverse problem and is ill-posed, meaning that there are many sets of
parameters that could give a similar quality of match, with no method to distinguish between sets for
accuracy without independent test sets that were not used as part of the fitting process. Figure 2 shows
the flow of data in a history matching process.
Noise Yes
Optimization Algorithim
Good
Calculated
Data Input Reservoir Model Match System Output
Output
?
No
New Model Parameters
History
Matching
Algorithm
Figure 2. History matching flowchart with the iterative process to refine model predictions for optimization.
History matching may be performed manually or with computer assistance. Before computer based
optimization methods were widely available, history matches were done manually; in other words,
an engineer would select the parameters based on a geologic model, run a simulation, and check
the quality of fit. Then, the entire process would be repeated with different parameters selected to
improve the fit. Such calculations were extremely time-consuming, with some simulations requiring
up to months of clock time. Automatic or assisted history matching, however, is now commonplace.
In automatic history matching, an optimization algorithm is used to rapidly run several simulations
while changing the parameters to find the best fit. There are many different optimization algorithms
that are commonly used and they can be broadly categorized into global or local algorithms.
Local algorithms find one candidate solution to the optimization problem that depends on the starting
initial guess and may not be the global optimum. Global algorithms may find several local solutions
or the global solution [43]. This section gives a brief overview of applications and common methods of
assisted history matching since 2010 and reviews the application of history matching applied to EOR.
For further discussion, the reader is referred to the review by [22,44].
Stochastic methods vary uncertain input parameters from a probability distribution function
to create an ensemble of model predictions or optimization results [45,46]. The predictions that
most closely align with the measurements are selected to find an optimal solution to the history
matching problem. Assisted history matching with stochastic methods may use evolutionary, particle
swarm, or simulated annealing algorithms. Stochastic methods search the solution space with some
degree of randomness and gradually refine the history match to greater accuracy. Most current
research developments involve modifications on these algorithms to improve accuracy, reliability, or
computation time. Rahmati et al. [47]. enhanced the differential evolution optimization algorithm to
improve convergence rates and robustness. This algorithm was compared to a gradient-based method
and a Genetic Algorithm (GA). In the study, the gradient-based method was only able to find local
minima due to proximity to the initial guess. The differential algorithm was found to converge with
nearly half the simulations of the GA [47]. Park et al. [48] extended a Pareto-based, Multi-Objective
Evolutionary Genetic Algorithm (MOEGA) developed by Deb et al. [49] to handle conflicting multiple
objectives to history matching [48]. This algorithm is significant because there are generally several
sets of data that are to be matched, such as production data, time-lapse seismic data, etc., each with
a unique objective function. Generally, these several objectives are lumped into one weighted-sum
function that is used for the history match. This can lead to incomplete solutions as the optimization of
one objective may increase the error of another. A Multi-Objective Evolutionary Algorithm (MOEA)
is able to create a set of solutions that deal with each objective separately and then works to find the
best combination. By creating a set of solutions using MOEA, the uncertainty can be measured to
some degree. Hajizadeh et al. [50] applied ant colony optimization (ACO) for history matching and
uncertainty quantification to both a simple and a complex reservoir model. They found that ACO
converged well and was able to perform better than a neighborhood algorithm in their tests.
A common objective in recent studies is to compare two or more methods using the same
reservoir data in order to find the best history matching algorithm. ACO, differential evolution (DE),
particle swarm optimization (PSO), and the neighborhood algorithm (NA) are all directly compared
by Hajizadeh et al. [51] . For a small, simple reservoir with one producing well, the history match of
PSO and NA were the best, with PSO achieving slightly better results. On a large synthetic reservoir
with complex geology and multiple wells, DE achieved the best history match, and NA was unable to
find a satisfactory match [51]. Three stochastic algorithms for joint history matching of production and
4D seismic data were compared by Jin et al. [52]. The methods were Very Fast Simulated Annealing
(VFSA), PSO, and NA. In two test fields, the researchers found that PSO, while being slightly more
computationally expensive, had good convergence characteristics and was more effective than the
other two methods. They also found that VFSA has certain advantages, but requires more iterations
and thus is generally more useful when more computational power is available [52]. Currently, there
is no method that clearly and consistently outperforms other methods for all reservoirs. While a single
Processes 2017, 5, 34 6 of 25
best solution for a reservoirs is unlikely, future research should be performed to determine the optimal
algorithm for different types of reservoirs and recovery methods.
An Ensemble Kalman Filter (EnKF) updates and predicts states and parameters of the model.
Much of the history matching research in recent years has been focused on the use of the EnKF.
The EnKF is “a sequential filter method, which means that the model is integrated forward in
time and, whenever measurements are available, these are used to reinitialize the model before the
integration continues” [53]. The greatest advantages of using EnKF for history matching are the ease
of implementation and ability to handle nonlinear systems. EnKF is also easily parallelizable, however
the computational cost is high compared to other filter methods. Aanonsen et al. [18] reviewed the use
of the EnKF in reservoir engineering from its inception in 1994 to 2009.
Most current developments in the use of EnKF deal with decreasing computation time, ensuring
geological consistency, or accurately matching strongly nonlinear reservoirs. A reduced order flow
model, in this case Trajectory Piecewise Linearization (TPWL) form, is used in order to increase the
ensemble size for the EnKF at low computational cost [54]. This avoids possible issues of ensemble
collapse while improving computation time. They demonstrated that using the EnKF on 50 full-fidelity
simulations and 150 TPWL simulations gave comparable results to using 200 full-fidelity simulations
and resulted in a better history match than using only 50 full-fidelity simulations. Heidari et al. [55]
combined the EnKF with pilot points and gradual deformation in order to prevent the EnKF from
departing from prior information and to prevent petrophysical properties from reaching non-physical
values. Elsheikh et al. [56] proposed an alternative to the EnKF called the iterative stochastic
ensemble method (ISEM). ISEM estimates gradients stochastically using an ensemble of gradients.
Agbalaka et al. [57] implemented a two-stage ensemble-based history matching technique that allows
the EnKF to history match problems that have strong nonlinearities and have multimodal probability
distribution functions. Their approach consisted of one EnKF to assimilate all production data followed
by a second EnKF that resamples the model variables around a mode. Using their approach, they
were able to greatly increase the accuracy of the history match. Emerick and Reynolds [58] created
a study using nine recently developed ensemble-based methods to compare quality of the data
match, uncertainty quantification, and computation time. The study was performed on a small
nonlinear reservoir model. They found that the ensemble smoother with multiple data assimilations
and the ensemble random maximum likelihood (EnRML) methods gave the best quality of match.
The ensemble smoother with multiple data assimilations also gave the best quantification of uncertainty.
The regular ensemble smoother had the lowest computation time, followed by deterministic EnKF and
then the normal EnKF. They also found that several EnKF methods were unable to find an acceptable
match, showing one of the major drawbacks of the EnKF. Adjusting the EnKF to allow accurate matches
for highly nonlinear reservoir models will likely be a main area for future research.
Another group of commonly used methods can be classified as gradient-based as opposed
to derivative-free methods. These methods rely on calculating derivatives of the model equations
in order to find the optimal set of parameters. They are generally very computationally efficient,
but often difficult to implement because implementation requires access to simulator source code.
Gradient-based methods often do not find the global solution. Common objectives for research
on gradient-based history matching in recent years include combining aspects of gradient-based
history matching with other techniques in order to improve accuracy and make implementation
easier. Ding [59] developed a data partition technique for gradient-based history matching. Using this
technique, they separated the objective function into local components, each of which depends on
fewer parameters, and then computed all the derivatives of these objective functions. This greatly
assists in the history match of very large fields. Kaleta et al. [60] proposed an approach
to gradient-based history matching that uses model reduction instead of a full adjoint-based
approach. This was shown to reduce the computational efficiency slightly but greatly simplify the
implementation. Bhark et al. [61] used gradient-based optimization within a framework of adaptively
scaled frequency-domain parameterization. This method reduces the number of parameters through
Processes 2017, 5, 34 7 of 25
cosine parameterization and uses gradient-based minimization on the data misfit. Then, the model is
gradually refined to include more parameters and the minimization is repeated. This whole process
continues until no further improvements occur. The advantage of this method is that it avoids local
minima while maintaining stability.
Table 1. Comparison of history matching techniques. A (+) indicates a benefit or positive feature and a
(−) indicates a drawback or limitation of each method.
Table 1 provides a comparison of the different history matching algorithms [22]. There is not one
algorithm that clearly outperforms others when history matching in waterflood or EOR situations.
For fields under waterflood or EOR, where reliable geologic data is difficult to obtain, reduced
order proxy modeling presents a method for determining the relationship between controllable
reservoir system inputs and oil production. Various data driven models have been applied to the
history matching problem including finite impulse response [79], autoregressive models, [36,37,42],
and the capacitance resistance model [38–41]. These models have been coupled with various history
matching algorithms including gradient based methods [38] and EKF [80]. One of the major advantages
of using reduced order proxy models is the reduced computation time. These models can also quickly
provide an understanding of well interactions, and may even be used to infer some geologic structures.
However these models lack the ability to incorporate geologic data into the model, and may have difficulty
modeling very nonlinear reservoirs, making them less robust than larger, more complex models.
The top down or hierarchical approach to history matching has been applied with success to
various reservoirs. The top down approach uses the simplest model necessary for the business decision
at hand. Williams et al. [81] describe an algorithm for performing manual history matching that
uses a top down approach, first matching reservoir pressure and flow rates followed by tuning of
the individual reservoir layers and wells, in a structured manner. This idea of a top down approach
was extended to automatic history matching in [82]. The top down approach allows for various
global matches to subsequently be matched on the local level, providing a good understanding of
uncertainty when compared to building a single detailed model for which to perform history matching.
This method has been applied to various real fields with success [82–84]. While the top down approach
Processes 2017, 5, 34 9 of 25
allows for easier understanding of reservoir uncertainties, solving the global and local problems
separately may not reveal all possible history matches.
δx
s.t. 0 = f ( , y, x, p, u) (3b)
δt
0 = g(y, x, p, u) (3c)
0 ≤ h(y, x, p, u) (3d)
This section examines well placement, well control, and coupled well placement-well control
optimization. These types of optimization problems are often referred to as production optimization.
The solution methods include stochastic and adjoint techniques. Common stochastic methods include
GA, ensemble based optimization (EnOpt), particle swarm optimization (PSO), simulated annealing
(SA), and to a lesser extent harmony search methods. A gradient based adjoint method is common
as well.
for many stochastic or finite difference methods. Several researchers advocate the use of continuous
variables versus discrete variables so as to be able to more easily implement an adjoint method to
find optimal locations for wells [89–92]. This allows for wells to initially be placed in each grid block
and non-optimal well locations to be removed more readily [92]. Zhang et al. [92] further focus on
adjoint optimization for increasing overall NPV during waterflood. Large time steps (days or weeks)
are used to optimize the location of a specified number of water injection wells. When compared to
most stochastic methods fewer simulation runs are needed for this adjoint well placement method,
as only a forward and a reverse or adjoint simulation are required. In addition to adjoint optimization,
automatic differentiation poses a potential for improvements in reservoir history matching and
optimization [93,94]. Automatic differentiation is generally more accurate than numerical methods
and can more easily compute higher deviates and partial derivatives with respect to many inputs in
gradient based optimization algorithms.
Stochastic methods such as GA, PSO, SA, and harmony search have found use in the well
placement problem as their derivative free techniques allow for easy use of discrete information
common in well placement problems. While the adjoint method is generally more efficient at finding
optimums, it is difficult to implement and cannot be straightforwardly applied for discrete variables
such as well locations in a reservoir grid.
Evolutionary algorithms, which are perhaps the most common and well known stochastic
optimization algorithm group, have found merit in well placement optimization. Montes et al. [95]
propose that the nonlinearity and non-continuity common in well placement optimization limit
traditional simplex or gradient methods, and suggest GAs to effectively find optimal well locations.
Generally, binary GAs are used in well placement optimization, however continuous GAs are being
advocated by Farshi [96], Bukhamsin et al. [97] as they allow for smoother computation of optimal well
placement. Farshi [96] also expands the use of GAs by implementing a minimum Euclidian distance
between individuals within the population. Thus exploiting engineering intuition such as requiring a
minimum distance between wells. The inclusion of curved well placement in addition to straight wells
is also considered. Bukhamsin et al. [97] use continuous GAs to optimize the development of a real field
located in the Middle East and consider complicated well geometry such as multilateral wells (MLWs)
and maximum reservoir contact (MRC) wells. Ariadji et al. [85] use GAs in a commercial simulator
with varying well drainage radii and find this is a superior method to traditional guess and check.
Emerick et al. [98] utilize the Genocop III algorithm (GA for Numerical Optimization of Constrained
Problems) to handle the well placement constraints. The constraints include grid size, maximum
length of wells, minimum distance between wells, inactive grid cells, as well as user defined regions of
non-uniform shape where well placement is forbidden. The method applied by Emerick et al. [98] does
not utilize a proxy model and in some cases involves more than 100 decision variables and thousands
of reservoir simulation runs, possibly limiting the utility of the method. Bouzarkouna et al. [99] propose
optimal well location can be found through the use of a covariance matrix adaptation evolution
strategy that uses adaptive penalization with rejection to deal with constraints. This method is used in
conjunction with a meta-model that utilizes an approximate stochastic ranking procedure to rank the
fitness of individual wells.
Another stochastic technique employed for optimal well placement is PSO. PSO is a stochastic
optimization method based on the behavior of birds that sends particles throughout the solution space.
Every particles follows the optimum particle until the swarm of particles arrives at a solution [100].
Onwunalu and Durlofsky [101] apply PSO for determining well location and well type (vertical,
horizontal, deviated, multilateral, injector or producer). In this method, particles represent individual
wells with a random neighborhood topology being utilized and updated after each iteration with an
absorb technique to handle infeasible particles. In the special case of invalid deviated or multilateral
well configurations, such as intersecting wells, these problems are resolved through assignment to
a negative objective function and use of a penalty method. Cheng et al. [102] use a niche particle
swarm algorithm developed by [103] to optimize the well placement of production wells. The niche
Processes 2017, 5, 34 11 of 25
particle swarm method divides the swarm into smaller subgroups, which then search for the optimal
solution of that subgroup. This method increase diversity of the swarm and improves the probability
of finding a global maximum when compared to PSO. Nwankwor et al. [104] advocate use of a
hybrid differential evolution and particle swarm algorithm to find optimal well locations for a
specified number of wells and maintain that the hybrid method employed is more efficient than
an evolutionary or PSO performed separately. Ding et al. [105] investigate the use of different PSOs
for the optimal well placement problem. The different PSOs utilized include modified PSO, standard
PSO, centered-progressive PSO, and modified PSO and quality map method. This study shows that
the modified PSO and quality map method are better able to optimally place wells than the other three
methods, while modified PSO outperforms simple PSO and center-progressive PSO methods.
Genetic algorithms (GA) are also applied to well placement optimization. Min et al. [106] find
the optimal location and number of waterflood injection wells through the use of a multi-objective
GA. This multi-objective GA is based on non-dominated sorting and crowding distance sorting, which
allows for costs to be minimized and revenue maximized. Hybrid GAs are finding merit in reservoir
optimization. Like GAs, Simulated Annealing (SA) has been applied in well placement optimization.
SA is modeled after the annealing of metals, where the material is heated and then allowed to slowly
cool. Similarly, in SA solutions are selected that lower the objective function while a certain probability
of higher solutions are chosen as well. This prevents the algorithm from falling into local minimum.
Beckner and Xong [107] present well placement and scheduling as a traveling salesman problem
and propose the idea that the ability of SA to deduce non-uniform well spacings makes it a viable
optimization technique. Results indicate that non uniform well spacing is optimal and that optimal
well spacing results in significant increases in field development profitability [107]. A variant of SA
known as VFSA is also finding use in the well placement problem [108]. VFSA constrains the search
space as the algorithm progresses, thereby making the algorithm more likely to accept steps that
reduce the objective function. Bangerth et al. [109] find that in two different test cases VFSA performs
best at finding the global optimum in the well placement problem when compared to simultaneous
perturbation stochastic approximation, finite difference gradient, the Nelder-Mead simplex, and a GA,
although VFSA often requires more function evaluations than the simultaneous perturbation stochastic
approximation method.
To a small extent harmony search is used in the well placement problem. Afshari et al. [110]
propose a method for well placement optimization using streamline simulation with an improved
harmony search (IHS) algorithm. The IHS method employed is a simple method with few parameters
that performs well when compared to other stochastic algorithms [110].
injection rates of individual water injection wells to be suitably altered. Li-xin and Jian-jun [114] assert
that a hybrid GA minimizes injection energy consumption while maintaining a desired production in a
high pressure waterflood. Mamghaderi and Pourafshary [40] use a GA with the CRM to allocate water
volumes amongst injectors and maximize oil production. Safarzadeh et al. [115] use a multi-objective
GA with a streamline simulation. They maintain that this combination allows for higher oil production,
lower water cut, and less water lost to aquifers than traditional methods.
SA is being pursued to some extent in waterflooding optimization. Yang et al. [116] argue that both
SA and a GA can be employed to optimize and control the nonlinear production-injection operation
system for water and gas flooding. This technique is shown to slow down the flooding front in a
stable manner. Further, Khan et al. [117] show that short term oil rates and long term recovery can be
maximized in waterfloods by using constrained SA with a full field reservoir simulator. Constraints in
this method are placed on the voidage replacement ratio, reservoir pressure, sweep efficiencies, and
production and injection rates.
PSO is also finding some use in waterflooding optimization, but has not been studied extensively.
Wang et al. [118] investigate the use of the ant colony particle swarm and find that the algorithm is
effectively able to optimize waterflooding injection rates.
Well control optimization may also be performed using reduced order proxy models. As mentioned
in Section 2.1, reduced order proxy models present a method for determining the relationship between
injection and production wells. With these relationships, injection optimization can be performed
using gradient or stochastic methods. These methods are particularly useful for large fields where
high fidelity modeling would be computationally expensive. Various reduced order proxy models
have been applied to fields under waterflooding or EOR. Weber [38] used the capacitance resistance
model to optimize NPV at various oil prices using a gradient based solver. Other models such as
finite impulse response and auto regressive models [42] that have been used for history matching
injector producer relationships could be optimized with gradient based solvers or stochastic methods.
Hourfar et al. [42] apply system identification techniques from the process control field and treat the
reservoir as a linear multiple input multiple output (MIMO) process. Their results show simultaneous
handling of reservoir nonlinearities and time-varying reservoir conditions with linear models that
continuously adapt based on recent data. These simple models can easily be optimized, however their
simplicity reduces the confidence in extrapolating these models into the future. Often these models are
only valid for shorter time scales and must be refit with new data, making longer scale optimization
more difficult with these models.
During well control optimization, determining the size of each control step is normally not
performed a priori. This practice can lead to problems with too many or too few degrees of freedom.
With too many control steps, optimization algorithms may find various equivalent solutions while
computation times are higher. With too few time steps, the optimal solution may not be possible
or misleading. Oliveira et al. [119] developed a method to determine the optimal number and size
of control steps for well control optimization. The results show improvements in NPV while also
decreasing computation time.
In addition to optimal injection flowrates to increase production pressure, artificial lift is often
installed to supplement when there is low production pressure. The pump liquid level is detected
throughout cyclic load and position trends referred to as downhole cards [120,121]. A liquid level is
detected if the load shifts part way through the down cycle of a rod pump and adjustments can be
made to maximize pump fillage [122]. By actively monitoring the liquid level, injector rates nearby can
be increased or decreased to maintain a certain target level in the production well annulus. A future
area of research is to coordinate injector action with pump-off action using variable stroke length and
speed of rod pumping [123] to increase the reservoir NPV [124].
Processes 2017, 5, 34 13 of 25
1. A GA can be applied with a high fidelity reservoir simulator, although this may require excessive
computational time.
2. GAs can be applied to a surrogate model to reduce computational time. In this instance a surrogate
model, such as an adaptive ANN, is used with an intelligent sample selection algorithm derived
from a simulator. The GA is used to optimize the surrogate model.
Processes 2017, 5, 34 14 of 25
3. A surrogate model is updated after each generation of the GA based on optimal solutions from
the Pareto front.
Particle swarm has also shown some effectiveness in coupled placement and rate optimization.
Isebor and Durlofsky [135], Isebor et al. [136] propose that a hybrid particle swarm-mesh adaptive
direct search method is adequately able to optimize well location, well controls, and number of wells.
This method filters for nonlinear constraint handling, along with a bi-objective optimization that seeks
to minimize the objective function along with constraint violation, and is shown to outperform the
sequential optimization of well location and injection rate.
Stochastic and gradient methods have been used extensively to solve both well placement and
injection optimization problems. Table 2 provides an overview of the advantages and disadvantages
of the optimization methods reviewed.
Table 2. Comparison of methods for production optimization. A (+) indicates a benefit or positive
feature and a (−) indicates a drawback or limitation of each method.
hybrid genetic-particle swarm algorithm to maximize NPV by holding slug concentrations constant
and using control vector parameterization with constant parameters obtained through nonlinear
programming. This hybrid algorithm, it is argued, is able to more quickly converge to a solution
than a GA done independently through using the PSO position displacement strategy as a mutation
operator [141]. Horowitz et al. [155] present a method of maximizing NPV by using a two stage method
that utilizes an efficient global optimization algorithm [156] to find the optimal value for concentration
of polymer in water, as well as duration of slugs and time at which to begin injections for individual
injectors. This optimization is accomplished by first using a Latin hypercube technique to obtain an
initial sample of designs, after which simulation runs are used to construct a metamodel. Thereafter, infill
sampling criterion are used, in which the metamodel is used to lead the search for promising designs to
be added to the sample to update the metamodel until desired criteria are met. Hui et al. [142] propose
that a particle swarm algorithm to minimize water cut while maximizing production to find optimal
polymer concentrations. Lei et al. [148] propose a hybrid GA that uses the position displacement
method of particle swarm to solve a mixed integer optimization problem, which seeks to find optimal
slug size and polymer concentration. Ekkawong [149] implements a multi-objective GA to optimize
oil production and polymer utility function.
Various authors have examined optimizing production of steam-over-solvent processes for
heavy oil reservoirs, looking specifically at injection rate and solvent concentrations as adjustable
parameters to optimize NPV. Gates and Chakrabarty [145] utilize both GA and SA to optimize
NPV for Steam Assisted Gravity Drainage (SAGD). Solvent additive for SAGD is optimized.
Al-Gosayir et al. [150] examine the use of a hybrid GA in heterogeneous reservoirs. They investigate
optimizing steam-over-solvent processes for fractured carbonate reservoirs by using a GA with a
near-orthogonal array to better seed initial populations and a proxy model run to find the optimal
injection rates and concentrations for cyclic injection. Zhao et al. [146] show that simulated annealing
finds optimum pressure and temperature of injected hot water for recovery of thin heavy oil in various
stages of reservoir depletion.
CO2 EOR is increasingly an important area of interest as a potential final use for carbon capture
and sequestration initiatives [157,158]. Emera and Sarma [151] propose a GA to solve the minimum
miscibility CO2 pressure needed for desired oil recovery. The proposed method uses a roulette wheel
selection technique. Furthermore, Dehghani et al. [152] propose that a hybrid neural GA is useful
at predicting minimum miscibility pressure needed for EOR. Ali Ahmadi et al. [143] argue that a
hybrid artificial neural network and particle swarm algorithm are able determine the minimum
miscible pressure. This method is better able to predict production in various thermodynamic
environments [144]. Chen et al. [153] optimized a water-alternating-gas process using a hybrid
GA. Little research has been devoted to adjoint methods in gas flooding optimization. Mehos and
Ramirez [140] show that an adjoint method is effective at finding the minimum miscible CO2 pressure
needed for optimal oil production.
4. Benchmark Models
Benchmark models have been developed to facilitate the comparison of different optimization
methods. A benchmark model gives multiple researchers the exact same data and starting point.
Thus the data produced by these groups can be compared more accurately. This section describes
different benchmark models that have been provided by several organizations to facilitate such
comparison, giving a brief review of three benchmarks, their properties, and the uses of each.
required reservoir properties. Historical data for the study was created using an upscaled version of the
model containing 450,000 cells, and many participants used a model further upscaled to 60,000 cells to
perform history matching and optimization. The reservoir is under-saturated with 20 producing wells
and 10 injectors. Additional properties can be found in the paper by Peters et al. [62]. The Benchmark
provides enough data to run and compare various history matching and optimization methods.
From the competition/workshop the groups involved used history matching and optimization
techniques to optimize production with respect to NPV. These results were compared to the known
solution. The truth case realized a NPV of 4.66 × 109 USD. The highest realized NPV from the
OU/Chevron group, and was 4.53 × 109 USD, while the lowest was 4.19 × 109 USD. The OU/Chevron
group maximized the NPV with liquid rates as control variables using EnOpt and also used EnRML for
their history matching strategy [159]. A detailed summary of the results is given by Peters et al. [62].
Table 3. Summary of SPE benchmark problems for reservoir simulation, history matching,
and optimization.
Acknowledgments: This work was performed with support from the International Reservoir Simulation Research
Institute (IRSRI) at Brigham Young University (BYU). The authors would like to acknowledge the helpful guidance
from Hugh Hales, founder of IRSRI, and several others who have contributed to the review.
Processes 2017, 5, 34 18 of 25
References
1. Enhanced Oil Recovery. Available online: http://energy.gov/fe/science-innovation/oil-gas-research/
enhanced-oil-recovery (accessed on 16 May 2017).
2. Mashayekhizadeh, V.; Kord, S.; Dejam, M. EOR potential within Iran. Spec. Top. Rev. Porous Media Int. J.
2014, 5, 325–354.
3. Mollaei, A.; Maini, B. Steam flooding of naturally fractured reservoirs: Basic concepts and recovery
mechanisms. J. Can. Pet. Technol. 2010, 49, 65–70.
4. Baker, P. An experimental study of heat flow in steam flooding. Soc. Pet. Eng. J. 1969, 9, 89–99.
5. Akstinat, M. Gas evolution and change of oil composition during steam flooding of oil reservoirs. J. Pet. Geol.
1983, 5, 363–388.
6. Yitang, Z.; Xiuluan, L.; Xia, Z. Four fundamental principles for design and follow-up of steam flooding in
heavy oil reservoirs. Pet. Explor. Dev. 2008, 35, 715–719.
7. Saboorian-Jooybari, H.; Dejam, M.; Chen, Z. Heavy oil polymer flooding from laboratory core floods to pilot
tests and field applications: Half-century studies. J. Pet. Sci. Eng. 2016, 142, 85–100.
8. Saboorian-Jooybari, H.; Dejam, M.; Chen, Z. Half-Century of Heavy Oil Polymer Flooding from Laboratory
Core Floods to Pilot Tests and Field Applications. In Proceedings of the SPE Canada Heavy Oil Technical
Conference, Calgary, AB, Canada, 9–11 June 2015; Society of Petroleum Engineers: Calgary, AB, Canada, 2015.
9. Guo, Y.; Liang, Y.; Cao, M.; Feng, R.; Zhang, X.; Li, H.; Hu, J. Flow Behavior and Viscous-Oil-
Microdisplacement Characteristics of Hydrophobically Modified Partially Hydrolyzed Polyacrylamide
in a Repeatable Quantitative Visualization Micromodel. SPE J. 2017, doi:10.2118/185185-PA.
10. Seright, R.S. How Much Polymer Should Be Injected During a Polymer Flood? Review of Previous and
Current Practices. SPE J. 2016, doi:10.3997/2214-4609.201700317.
11. Seright, R. How Much Polymer Should Be Injected During a Polymer Flood? In Proceedings of the SPE
Improved Oil Recovery Conference, Tulsa, OK, USA, 11–13 April 2016; Society of Petroleum Engineers:
Tulsa, OK, USA, 2016.
12. Liu, S.; Zhang, D.; Yan, W.; Puerto, M.; Hirasaki, G.J.; Miller, C.A. Favorable attributes of alkaline–
surfactant–polymer flooding. SPE J. 2008, 13, 5–16.
13. Zerpa, L.E.; Queipo, N.V.; Pintos, S.; Salager, J.L. An optimization methodology of alkaline–surfactant–
polymer flooding processes using field scale numerical simulation and multiple surrogates. J. Pet. Sci. Eng.
2005, 47, 197–208.
14. Samimi, S.E.; Masihi, M.; Gerami, S.; Ghazvini, M.G. An Improvement on Modeling of Forced Gravity
Drainage in Dual Porosity Simulations Using a New Matrix-Fracture Transfer Function. Transp. Porous Media
2012, 94, 207–223.
15. Lin, L.; Yang, Z.; Liu, X.; Sun, J.; He, Y.; Guo, H. Experiments on the permeability limits of tight oil reservoirs
for gas flood recovery. Spec. Top. Rev. Porous Media Int. J. 2016, 7, 385–390.
16. Jansen, J. Model-based control of subsurface flow. J. Process Control 2008, 18, 846–855.
17. Jansen, J.D.; Durlofsky, L.J. Use of reduced-order models in well control optimization. Optim. Eng. 2016, 18,
105–132.
18. Aanonsen, S.I.; Nævdal, G.; Oliver, D.S.; Reynolds, A.C.; Vallès, B. The ensemble Kalman filter in reservoir
engineering—A review. SPE J. 2009, 14, 393–412.
19. Alvarado, V.; Manrique, E. Enhanced oil recovery: An update review. Energies 2010, 3, 1529–1575.
20. Franco, A.; Vaccaro, M. Numerical simulation of geothermal reservoirs for the sustainable design of energy
plants: A review. Renew. Sustain. Energy Rev. 2014, 30, 987–1002.
21. Islam, A.W.; Sepehrnoori, K. A Review on SPE’s Comparative Solution Projects (CSPs). J. Pet. Sci. Res. 2013,
2, doi:10.14355/jpsr.2013.0204.04.
22. Oliver, D.S.; Chen, Y. Recent progress on reservoir history matching: A review. Comput. Geosci. 2011,
15, 185–221.
23. Jansen, J. Adjoint-based optimization of multi-phase flow through porous media—A review. Comput. Fluids
2011, 46, 40–51.
Processes 2017, 5, 34 19 of 25
24. Hou, J.; Zhou, K.; Zhang, X.S.; Kang, X.D.; Xie, H. A review of closed-loop reservoir management. Pet. Sci.
2015, 12, 114–128.
25. Beal, L.D.; Park, J.; Petersen, D.; Warnick, S.; Hedengren, J.D. Combined model predictive control and
scheduling with dominant time constant compensation. Comput. Chem. Eng. 2017, 104, 271–282.
26. Safdarnejad, S.M.; Hedengren, J.D.; Baxter, L.L. Plant-level dynamic optimization of cryogenic carbon
capture with conventional and renewable power sources. Appl. Energy 2015, 149, 354–366.
27. Safdarnejad, S.M.; Hedengren, J.D.; Baxter, L.L. Dynamic optimization of a hybrid system of energy-storing
cryogenic carbon capture and a baseline power generation unit. Appl. Energy 2016, 172, 66–79.
28. Powell, K.M.; Eaton, A.N.; Hedengren, J.D.; Edgar, T.F. A continuous formulation for logical decisions in
differential algebraic systems using mathematical programs with complementarity constraints. Processes
2016, 4, 7.
29. Safdarnejad, S.M.; Hedengren, J.D.; Lewis, N.R.; Haseltine, E.L. Initialization strategies for optimization of
dynamic systems. Comput. Chem. Eng. 2015, 78, 39–50.
30. Udy, J.; Blackburn, L.; Hedengren, J.; Darby, M. Reduced Order Modeling for Reservoir Injection
Optimization and Forecasting. In Proceedings of the FOCAPO/CPC Conference, Tuscon, AZ, USA,
8–12 January 2017; Foundations of Computer Aided Process Operations/Chemical Process Control: Tuscon,
AZ, USA, 2017.
31. Todd, M.; Longstaff, W. The development, testing, and application of a numerical simulator for predicting
miscible flood performance. J. Pet. Technol. 1972, 24, 874–882.
32. Wang, P.; Yotov, I.; Wheeler, M.; Arbogast, T.; Dawson, C.; Parashar, M.; Sepehrnoori, K. A New Generation
EOS Compositional Reservoir Simulator: Part I-Formulation and Discretization. In Proceedings of the SPE
Reservoir Simulation Symposium, Dallas, TX, USA, 8–11 June 1997; Society of Petroleum Engineers: Dallas,
TX, USA, 1997.
33. Kazemi, H.; Vestal, C.; Shank, D.G. An efficient multicomponent numerical simulator. Soc. Pet. Eng. J. 1978,
18, 355–368.
34. Blunt, M.J.; Jackson, M.D.; Piri, M.; Valvatne, P.H. Detailed physics, predictive capabilities and macroscopic
consequences for pore-network models of multiphase flow. Adv. Water Resour. 2002, 25, 1069–1089.
35. Lee, K.; Ortega, A.; Jafroodi, N.; Ershaghi, I. A multivariate autoregressive model for characterizing producer
producer relationships in waterfloods from injection/production rate fluctuations. SPE West. Reg. Meet.
2010, 13, 1–11.
36. Liu, F.; Mendel, J.M.; Nejad, A.M. Forecasting injector/producer relationships from production and injection
rates using an extended Kalman filter. SPE J. 2009, 14, 653–664.
37. Zhai, D.Y.; Mendel, J.M. An inequality-constrained extended Kalman filter for continual forecasting of
interwell connectivities in waterfloods. SPE West. Reg. Meet. 2010, doi:10.2118/134006-MS.
38. Weber, D. The use of Capacitance-Resistance Models to Optimize Injection Allocation and Well Location in
Water Floods. Ph.D. Thesis, The University of Texas at Austin, Austin, TX, USA, 2009.
39. Sayarpour, M.; Zuluaga, E.; Kabir, C.S.; Lake, L.W. The use of capacitance–resistance models for rapid
estimation of waterflood performance and optimization. J. Pet. Sci. Eng. 2009, 69, 227–238.
40. Mamghaderi, A.; Pourafshary, P. Water flooding performance prediction in layered reservoirs using improved
capacitance-resistive model. J. Pet. Sci. Eng. 2013, 108, 107–117.
41. Jafroodi, N.; Zhang, D. New method for reservoir characterization and optimization using CRM–EnOpt
approach. J. Pet. Sci. Eng. 2011, 77, 155–171.
42. Hourfar, F.; Moshiri, B.; Salahshoor, K.; Zaare-Mehrjerdi, M.; Pourafshary, P. Adaptive modeling of
waterflooding process in oil reservoirs. J. Pet. Sci. Eng. 2016, 146, 702–713.
43. Verga, F.; Cancelliere, M.; Viberti, D. Improved application of assisted history matching techniques. J. Pet.
Sci. Eng. 2013, 109, 327–347.
44. Rwechungura, R.W.; Dadashpour, M.; Kleppe, J. Advanced History Matching Techniques Reviewed.
In Proceeding of the SPE Middle East Oil and Gas Show and Conference, Manama, Bahrain, 25–28 September
2011; Society of Petroleum Engineers: Manama, Bahrain, 2011.
45. Stordal, A.S.; Szklarz, S.P.; Leeuwenburgh, O. A theoretical look at ensemble-based optimization in reservoir
management. Math. Geosci. 2016, 48, 399–417.
46. Anifowose, F.A.; Labadin, J.; Abdulraheem, A. Ensemble machine learning: An untapped modeling
paradigm for petroleum reservoir characterization. J. Pet. Sci. Eng. 2017, 151, 480–487.
Processes 2017, 5, 34 20 of 25
47. Rahmati, H.; Nouri, A.; Pishvaie, M.R.; Bozorgmehri, R. A modified differential evolution optimization
algorithm with random localization for generation of best-guess properties in history matching.
Energy Sources Part A Recovery Util. Environ. Effects 2011, 33, 845–858.
48. Park, H.Y.; Datta-Gupta, A.; King, M.J. Handling conflicting multiple objectives using Pareto-based
evolutionary algorithm during history matching of reservoir performance. J. Pet. Sci. Eng. 2014, 125, 48–66.
49. Deb, K.; Pratap, A.; Agarwal, S.; Meyarivan, T. A fast and elitist multiobjective genetic algorithm: NSGA-II.
IEEE Trans. Evol. Comput. 2002, 6, 182–197.
50. Hajizadeh, Y.; Christie, M.; Demyanov, V. Ant colony optimization for history matching and uncertainty
quantification of reservoir models. J. Pet. Sci. Eng. 2011, 77, 78–92.
51. Hajizadeh, Y.; Demyanov, V.; Mohamed, L.; Christie, M. Comparison of evolutionary and swarm intelligence
methods for history matching and uncertainty quantification in petroleum reservoir models. In Intelligent
Computational Optimization in Engineering; Springer: Berlin/Heidelberg, Germany, 2010; pp. 209–240.
52. Jin, L.; van den Hoek, P.J.; Alpak, F.O.; Pirmez, C.; Fehintola, T.; Tendo, F.; Olaniyan, E.E. A comparison of
stochastic data-integration algorithms for the joint history matching of production and Time-Lapse-Seismic
data. SPE Reserv. Eval. Eng. 2012, 15, 498–512.
53. Evensen, G. The Ensemble Kalman Filter: Theoretical formulation and practical implementation. Ocean Dyn.
2003, 53, 343–367.
54. He, J.; Sarma, P.; Durlofsky, L.J. Reduced-order flow modeling and geological parameterization for
ensemble-based data assimilation. Comput. Geosci. 2013, 55, 54–69.
55. Heidari, L.; Gervais, V.; Ravalec, M.L.; Wackernagel, H. History matching of petroleum reservoir models by
the ensemble Kalman Filter and parameterization methods. Comput. Geosci. 2012, 55, 84–95.
56. Elsheikh, A.H.; Wheeler, M.F.; Hoteit, I. An iterative stochastic ensemble method for parameter estimation
of subsurface flow models. J. Comput. Phys. 2013, 242, 696–714.
57. Agbalaka, C.C.; Stern, D.; Oliver, D.S. Two-stage ensemble-based history matching with multiple modes in
the objective function. Comput. Geosci. 2013, 55, 28–43.
58. Emerick, A.A.; Reynolds, A.C. Investigation of the sampling performance of ensemble-based methods with
a simple reservoir model. Comput. Geosci. 2013, 17, 325–350.
59. Ding, D.Y. Development of a data partition technique for gradient-based optimization methods in history
matching. SPE J. 2011, 16, 582–593.
60. Kaleta, M.P.; Hanea, R.G.; Heemink, A.W.; Jansen, J.D. Model-reduced gradient-based history matching.
Comput. Geosci. 2010, 15, 135–153.
61. Bhark, E.; Jafarpour, B.; Datta-Gupta, A. An adaptively scaled frequency-domain parameterization for
history matching. J. Pet. Sci. Eng. 2011, 75, 289–303.
62. Peters, L.; Arts, R.; Brouwer, G.; Geel, C.; Cullick, S.; Lorentzen, R.; Chen, Y.; Dunlop, N.; Vossepoel, F.;
Xu, R.; et al. Results of the Brugge benchmark study for flooding optimization and history matching.
SPE Reserv. Eval. Eng. 2010, 13, 391–405.
63. Naevdal, G.; Johnsen, L.M.; Aanonsen, S.I.; Vefring, E.H. Reservoir monitoring and continuous model
updating using ensemble Kalman filter. SPE J. 2005, 10, 66–74.
64. Bianco, A.; Cominelli, A.; Dovera, L.; Nævdal, G.; Valles, B. History matching and production forecast
uncertainty by means of the ensemble Kalman filter: A real field application. In Procedings of the
EUROPEC/EAGE Conference and Exhibition, London, UK, 11–14 June 2007; Society of Petroleum Engineers:
London, UK, 2007.
65. Jansen, J.D.; Brouwer, R.; Douma, S.G. Closed Loop Reservoir Management. In Proceedings of the SPE
Reservoir Simulation Symposium, The Woodlands, TX, USA, 2–4 February 2009; Society of Petroleum
Engineers: The Woodlands, TX, USA, 2009.
66. Skjervheim, J.; Evensen, G.; Aanonsen, S.; Rudd, B.; Johansen, T. Incorporating 4D Seismic Data in Reservoir
Simulation Models Using Ensemble Kalman Filter. In Proceedings of the SPE Annual Technical Conference
and Exhibition, Dallas, TX, USA, 9–12 October 2005; pp. 1–11.
67. Gu, Y.; Oliver, D. An iterative ensemble Kalman filter for multiphase fluid flow data assimilation. SPE J.
2007, 12, 438 – 446.
68. Cullick, A.S.; Johnson, W.D.; Shi, G. Improved and More Rapid History Matching with a Nonlinear Proxy
and Global Optimization. In Proceedings of the SPE Annual Technical Conference and Exhibition; Society of
Petroleum Engineers: San Antonio, TX, USA, 2006.
Processes 2017, 5, 34 21 of 25
69. Webb, S.J.; Bayless, J.S.; Dunlop, K.N.B. Enabling the ’Big Loop’-Ensuring Consistency of Geological and
Reservoir Simulation Models. In Proceedings of the AAPG Annual Convention and Exhibition, Long Beach,
CA, USA, 1–4 April 2007; pp. 1–4.
70. Sarma, P.; Durlofsky, L.J.; Aziz, K.; Chen, W.H. Efficient real-time reservoir management using adjoint-based
optimal control and model updating. Comput. Geosci. 2006, 10, 3–36.
71. Hooke, R.; Jeeves, T.A. Direct search solution of numerical and statistical problems. J. ACM 1961, 8, 212–229.
72. Echeverría Ciaurri, D.; Isebor, O.; Durlofsky, L. Application of derivative-free methodologies to generally
constrained oil production optimization problems. Proced. Comput. Sci. 2010, 1, 1301–1310.
73. He, Z.; Yoon, S.; Datta-Gupta, A. Streamline-based production data integration with gravity and changing
field conditions. SPE J. 2002, 7, 423–436.
74. Krymskaya, M.V.; Hanea, R.G.; Verlaan, M. An iterative ensemble Kalman filter for reservoir engineering
applications. Comput. Geosci. 2009, 13, 235–244.
75. Chen, Y.; Oliver, D.S. Levenberg-Marquardt forms of the iterative ensemble smoother for efficient history
matching and uncertainty quantification. Comput. Geosci. 2013, 17, 689–703.
76. Raniolo, S.; Dovera, L.; Cominelli, A.; Callegaro, C.; Masserano, F. History Match and Polymer Injection
Optimization in a Mature Field Using the Ensemble Kalman Filter. In Proceedings of the IOR 2013—17th
European Symposium on Improved Oil Recovery, St. Petersburg, Russia, 16–18 April 2013.
77. Douma, S.G.; Abu-Shiekah, I.M.; Kindi, Z.Y. Full Field History Matching for Chemical Flooding with the
Adjoint Method. In Proceedings of the SPE Kuwait Oil and Gas Show and Conference, Kuwait City, Kuwait,
8–10 October 2013; Society of Petroleum Engineers: Kuwait City, Kuwait, 2013.
78. Xu, S.; Zhang, M.; Zeng, F.; Chan, C. Application of Genetic Algorithm (GA) in history matching of the Vapour
Extraction (VAPEX) heavy oil recovery process. Nat. Resour. Res. 2014, doi:10.1007/s11053-014-9255-7.
79. Lee, K.H.; Ortega, A.; Nejad, A.M.; Ershaghi, I. A Method for Characterization of Flow Units Between
Injection-Production Wells Using Performance Data. In Proceedings of the 2008 SPE Western Regional and
Pacific Section AAPG Joint Meeting, Bakersfield, CA, USA, 31 March–2 April 2008; pp. 419–446.
80. Zhai, D.Y.; Jerry, M.; Mendel, F.L. A New Method for Continual Forecasting of Interwell Connectivitiy in
Waterfloods Using an Extended Kalman Filter. SPE West. Reg. Meet. 2009, doi:10.2118/121393-MS.
81. Williams, M.; Keating, J.F.; Barghouty, M. The stratigraphic method: A structured approach to history
matching complex simulation models. SPE Reserv. Eval. Eng. 1998, 1, 169–176.
82. Williams, G.; Mansfield, M.; MacDonald, D.; Bush, M. Top-down Reservoir Modelling. In Proceedings of
the SPE Annual Technical Conference and Exhibition, Houston, TX, USA, 26–29 September 2004; Society of
Petroleum Engineers: Houston, TX, USA, 2004.
83. Yin, J.; Park, H.Y.; Datta-Gupta, A.; King, M.J.; Choudhary, M.K. A hierarchical streamline-assisted history
matching approach with global and local parameter updates. J. Pet. Sci. Eng. 2011, 80, 116–130.
84. Cheng, H.; Dehghani, K.; Billiter, T.C. A Structured Approach for Probabilistic-Assisted History Matching
Using Evolutionary Algorithms: Tengiz Field Applications. In Proceedings of the SPE Annual Technical
Conference and Exhibition, Denver, CO, USA, 1–24 September 2008; Society of Petroleum Engineers: Denver,
CO, USA, 2008.
85. Ariadji, T.; Sukarno, P.; Sidarto, K.A.; Soewono, E.; Riza, L.S.; David, K. Optimization of Vertical Well
Placement for Oil Field Development Based on Basic Reservoir Rock Properties using a Genetic Algorithm.
ITB J. Eng. Sci. 2012, 44, 106–127.
86. Fahim Forouzanfar, A.; Gaoming Li, S. A Two-Stage Well Placement Optimization Method Based on
Adjoint Gradient. In Proceedings of the SPE Annual Technical Conference and Exhibition, Florence, Italy,
20–22 September 2010.
87. Bellout, M.C.; Echeverría Ciaurri, D.; Durlofsky, L.J.; Foss, B.; Kleppe, J. Joint optimization of oil well
placement and controls. Comput. Geosci. 2012, 16, 1061–1079.
88. Zandvliet, M.; Handels, M.; Van Essen, G.; Brouwer, D.; Jansen, J.D. Adjoint-based well-placement
optimization under production constraints. SPE J. 2008, 13, 392–399.
89. Handels, M.; Zandvliet, M.; Brouwer, R.; Jansen, J.D. Adjoint-Based Well-Placement Optimization under
Production Constraints. In Proceedings of the SPE Reservoir Simulation Symposium; Society of Petroleum
Engineers: Houston, TX, USA, 2007.
90. Wang, C.; Li, G.; Reynolds, A.C. Optimal Well Placement for Production Optimization. In Proceedings of the
Eastern Regional Meeting; Society of Petroleum Engineers: Lexington, KY, USA, 2007.
Processes 2017, 5, 34 22 of 25
91. Sarma, P.; Chen, W.H.; Durlofsky, L.J.; Aziz, K. Production optimization with adjoint models under nonlinear
control-state path inequality constraints. SPE Reserv. Eval. Eng. 2008, 11, 326–339.
92. Zhang, K.; Li, G.; Reynolds, A.C.; Yao, J.; Zhang, L. Optimal well placement using an adjoint gradient. J. Pet.
Sci. Eng. 2010, 73, 220–226.
93. Younis, R.M.; Aziz, K.; Tchelepi, H.; Alonso, J.J. Modern Advances in Software and Solution Algorithms for
Reservoir Simulation; Stanford University: Stanford, CA, USA, 2011.
94. Li, X.; Zhang, D. A backward automatic differentiation framework for reservoir simulation. Comput. Geosci.
2014, 18, 1009–1022.
95. Montes, G.; Bartolome, P.; Udias, A.L. The Use of Genetic Algorithms in Well Placement Optimization.
In Proceedings of the SPE Latin American and Caribbean Petroleum Engineering Conference, Buenos Aires,
Argentina, 25–28 March 2011; Society of Petroleum Engineers: Buenos Aires, Argentina, 2001.
96. Farshi, M.M. Improving Genetic Algorithms for Optimum Well Placement. Master’s Thesis, Stanford
University, Stanford, CA, USA, 2008.
97. Bukhamsin, A.Y.; Farshi, M.M.; Aziz, K. Optimization of Multilateral Well Design and Location in a Real
Field Using a Continuous Genetic Algorithm. In Proceedings of the SPE/DGS Saudi Arabia Section Technical
Symposium and Exhibition, Al-Khobar, Saudi Arabia, 4–7 April 2010; SPE: Al-Khobar, Saudi Arabia, 2010.
98. Emerick, A.A.; Silva, E.; Messer, B.; Almeida, L.; Szwarcman, D.; Pacheco, M.; Vellasco, M. Well Placement
Optimization Using a Genetic Algorithm With Nonlinear Constraints. In Proceedings of the SPE Reservoir
Simulation Symposium, The Woodlands, TX, USA, 2–4 February 2009; SPE: The Woodlands, TX, USA, 2009.
99. Bouzarkouna, Z.; Ding, D.Y.; Auger, A. Partially Separated Meta-Models with Evolution Strategies for Well
Placement Optimization. In Proceedings of the 73rd EAGE Conference & Exhibition Incorporating SPE
EUROPEC, Vienna, Austria, 23–26 May 2011; pp. 1–9.
100. Eberhart, R.C.; Kennedy, J. A New Optimizer Using Particle Swarm Theory. In Proceedings of the Sixth
International Symposium on Micro Machine and Human Science, New York, NY, USA, 4–6 October 1995;
pp. 39–43.
101. Onwunalu, J.E.; Durlofsky, L.J. Application of a particle swarm optimization algorithm for determining
optimum well location and type. Comput. Geosci. 2010, 14, 183–198.
102. Cheng, G.; An, Y.; Wang, Z.; Zhu, K. Oil Well Placement Optimization Using Niche Particle Swarm
Optimization. In Proceedings of the 2012 Eighth International Conference on Computational Intelligence
and Security, Guangzhou, China, 17–18 November 2012; pp. 61–64.
103. Brits, R.; Engelbrecht, A.P.; Van den Bergh, F. A Niching Particle Swarm Optimizer. In Proceedings of
the 4th Asia-Pacific Conference on Simulated Evolution and Learning, Singapore, 18–22 November 2002;
pp. 692–696.
104. Nwankwor, E.; Nagar, A.; Reid, D. Hybrid differential evolution and particle swarm optimization for
optimal well placement. Comput. Geosci. 2013, 17, 249–268.
105. Ding, S.; Jiang, H.; Li, J.; Tang, G. Optimization of well placement by combination of a modified particle
swarm optimization algorithm and quality map method. Comput. Geosci. 2014, 18, 747–762.
106. Min, B.; Park, C.; Kang, J.; Ahn, T.; Chung, S.; Kim, S. Optimal Injector Placement Coupled Multi-objective
Genetic Algorithm with a Black-oil Simulator in Waterflooding Project. In Proceedings of the 73rd EAGE
Conference & Exhibition, Vienna, Austria, 23–26 May 2011.
107. Beckner, B.; Xong, X. Field Development Planning Using Simulated Annealing-Optimal Economic Well Scheduling
and Placement; Technical Report; Society of Petroleum Engineers (SPE), Inc.: Dallas, TX, USA, 1995.
108. Parashar, M.; Klie, H.; Catalyurek, U.; Kurc, T.; Bangerth, W.; Matossian, V.; Saltz, J.; Wheeler, M.F.
Application of grid-enabled technologies for solving optimization problems in data-driven reservoir studies.
Future Gener. Comput. Syst. 2005, 21, 19–26.
109. Bangerth, W.; Klie, H.; Wheeler, M.F.; Stoffa, P.L.; Sen, M.K. On optimization algorithms for the reservoir oil
well placement problem. Comput. Geosci. 2006, 10, 303–319.
110. Afshari, S.; Aminshahidy, B.; Pishvaie, M.R. Application of an improved harmony search algorithm in well
placement optimization using streamline simulation. J. Pet. Sci. Eng. 2011, 78, 664–678.
111. Kraaijevanger, J.; Egberts, P.J.P.; Valstar, J.R.; Buurman, H.W. Optimal Waterflood Design Using the
Adjoint Method. In Proceedings of the SPE Reservoir Simulation Symposium, Houston, TX, USA,
26–28 February 2007; Society of Petroleum Engineers: Houston, TX, USA, 2007.
Processes 2017, 5, 34 23 of 25
112. Liu, X.; Reynolds, A. Gradient-based multi-objective optimization with applications to waterflooding
optimization. Comput. Geosci. 2015, 20, 677–693.
113. Zangl, G.; Hermann, R. Waterflood Pattern Optimization Using Genetic Algorithms with Multi-tank
Material Balance. In Proceedings of the SPE Annual Technical Conference and Exhibition, Houston, TX,
USA, 26–29 September 2004; Society of Petroleum Engineers: Houston, TX, USA, 2004.
114. Guan, X.J.; Wei, L.X.; Yang, J.J. Optimization of operation plan for water injection system in oilfield using
hybrid genetic algorithm. Acta Pet. Sin. 2005, 3, 026.
115. Safarzadeh, M.A.; Motealleh, M.; Moghadasi, J. A novel, streamline-based injection efficiency enhancement
method using multi-objective genetic algorithm. J. Pet. Explor. Prod. Technol. 2015, 5, 73–80.
116. Yang, D.; Zhang, Q.; Gu, Y. Integrated optimization and control of the production-injection operation
systems for hydrocarbon reservoirs. J. Pet. Sci. Eng. 2003, 37, 69–81.
117. Khan, H.; Saputelli, L.A.; Ranjan, P.; Carvajal, G.A.; Knabe, S.P.; Wang, F. Multi-Objectives Constrained
Waterflood Optimization in Tight Carbonates. In Proceedings of the SPE Reservoir Characterization and
Simulation Conference and Exhibition, Abu Dhabi, UAE, 16–18 September 2013; Society of Petroleum
Engineers: Abu Dhabi, UAE, 2013.
118. Wang, Y.X.; Liu, C.L.; Wang, Y.F.; Zhou, S.H. Optimization in oilfield water injection system based on
algorithm of ant colony-particle swarm method. J. Daqing Pet. Inst. 2010, 2, 014.
119. Oliveira, D.F.; Reynolds, A. An adaptive hierarchical multiscale algorithm for estimation of optimal well
controls. SPE J. 2014, 19, 909–930.
120. Everitt, T.A.; Jennings, J. An improved finite-difference calculation of downhole dynamometer cards for
sucker-rod pumps. SPE Prod. Eng. 1992, 7, 121–127.
121. Gibbs, S. A review of methods for design and analysis of rod pumping installations. J. Pet. Technol. 1982,
34, 2–931.
122. Ehimeakhe, V.M. Calculating Pump Fillage for Well Control Using Transfer Point Location. In Proceedings
of the SPE Eastern Regional Meeting, Morgantown, WV, USA, 13–15 October 2010; Society of Petroleum
Engineers: Morgantown, WV, USA, 2010.
123. Krug, D.; Nelsen, S.; Allison, J. Regenerative Hydraulic Lift System. U.S. Patent 8,562,308, 22 October 2013.
124. Emerick, A.A.; Portella, R.C.M. Production Optimization With Intelligent Wells. In Proceedings of the
Latin American & Caribbean Petroleum Engineering Conference, Buenos Aires, Argentina, 15–18 April 2007;
Society of Petroleum Engineers: Buenos Aires, Argentina, 2007.
125. Humphries, T.D.; Haynes, R.D.; James, L.A. Simultaneous and sequential approaches to joint optimization
of well placement and control. Comput. Geosci. 2014, 18, 433–448.
126. Humphries, T.D.; Haynes, R.D. Joint optimization of well placement and control for nonconventional well
types. J. Pet. Sci. Eng. 2015, 126, 242–253.
127. Sun, Z.; Xin, Y.; Sun, Q.; Ma, R.; Zhang, J.; Lv, S.; Cai, M.; Wang, H. Numerical Simulation of the
Depressurization Process of a Natural Gas Hydrate Reservoir: An Attempt at Optimization of Field
Operational Factors with Multiple Wells in a Real 3D Geological Model. Energies 2016, 9, 714.
128. Drouven, M.G.; Grossmann, I.E. Multi-period planning, design and strategic models for long-term,
quality-sensitive shale gas development. AIChE J. 2016, doi:10.1002/aic.15174.
129. Drouven, M.G.; Cafaro, D.C.; Grossmann, I.E. Stochastic Programming Models for Optimal Shale Well
Development and Refracturing Planning under Uncertainty. AIChE J. 2016, doi:10.1002/aic.15804.
130. Tavallali, M.; Karimi, I.; Halim, A.; Baxendale, D.; Teo, K. Well placement, infrastructure design, facility
allocation, and production planning in multireservoir oil fields with surface facility networks. Ind. Eng.
Chem. Res. 2014, 53, 11033–11049.
131. Brouwer, D.; Jansen, J. Dynamic optimization of water flooding with smart wells using optimal control
theory. SPE J. 2004, 9, 391–402.
132. Sarma, P.; Aziz, K.; Durlofsky, L.J. Implementation of Adjoint Solution for Optimal Control of
Smart wells. In Proceedings of the SPE Reservoir Simulation Symposium, The Woodlands, TX, USA,
31 January–2 Feburary 2005; Society of Petroleum Engineers: The Woodlands, TX, USA, 2005.
133. Forouzanfar, F.; Reynolds, A.C.; Li, G. Optimization of the well locations and completions for vertical and
horizontal wells using a derivative-free optimization algorithm. J. Pet. Sci. Eng. 2012, 86, 272–288.
Processes 2017, 5, 34 24 of 25
134. Sefat, M.H.; Salahshoor, K.; Jamialahmadi, M.; Vahdani, H. The Development of Techniques for the
Optimization of Water-flooding Processes in Petroleum Reservoirs Using a Genetic Algorithm and Surrogate
Modeling Approach. Energy Sources Part A Recovery Util. Environ. Effects 2014, 36, 1175–1185.
135. Isebor, O.J.; Durlofsky, L.J. Biobjective optimization for general oil field development. J. Pet. Sci. Eng. 2014,
119, 123–138.
136. Isebor, O.J.; Durlofsky, L.J.; Ciaurri, D.E. A derivative-free methodology with local and global search for the
constrained joint optimization of well locations and controls. Comput. Geosci. 2014, 18, 463–482.
137. Lei, Y.; Li, S.; Zhang, X.; Zhang, Q.; Guo, L. Optimal Control of Polymer Flooding Based on Maximum
Principle. J. Appl. Math. 2012, 2012, 987975.
138. Van Doren, J.; Douma, S.G.; Wassing, L.B.M.; Kraaijevanger, J.; De Zwart, A.H. Adjoint-Based Optimization
of Polymer Flooding. In Proceedings of the SPE Enhanced Oil Recovery Conference, Kuala Lumpur,
Malaysia, 19–21 July 2011; Society of Petroleum Engineers: Kuala Lumpur, Malaysia, 2011.
139. Namdar Zanganeh, M.; Kraaijevanger, J.; Buurman, H.; Jansen, J.; Rossen, W. Adjoint-Based Optimization of
a Foam EOR Process. In Proceedings of the ECMOR 13th European Conference on the Mathematics of Oil
Recovery, Biarritz, France, 10–13 September 2012.
140. Mehos, G.J.; Ramirez, W.F. Use of Optimal Control Theory to Optimize Carbon Dioxide Miscible-Flooding
Enhanced Oil Recovery. J. Pet. Sci. Eng. 1989, 2, 247–260.
141. Shurong, L.; Yang, L.; Xiaodong, Z.; Qiang, Z. Optimal Control Solving of Polymer Flooding Based on a
Hybrid Genetic Algorithm. In Proceedings of the IEEE 2010 29th Chinese Control Conference (CCC), Beijing,
China, 29–31 July 2010; pp. 5194–5198.
142. Hui, Z.; Yang, L.; Lin, C. A quantitative mathematic model for polymer flooding water-cut variation.
Pet. Explor. Dev. 2010, 37, 737–742.
143. Ali Ahmadi, M.; Zendehboudi, S.; Lohi, A.; Elkamel, A.; Chatzis, I. Reservoir permeability prediction by
neural networks combined with hybrid genetic algorithm and particle swarm optimization. Geophys. Prospect.
2013, 61, 582–598.
144. Sayyad, H.; Manshad, A.K.; Rostami, H. Application of hybrid neural particle swarm optimization algorithm
for prediction of MMP. Fuel 2014, 116, 625–633.
145. Gates, I.; Chakrabarty, N. Optimization of steam assisted gravity drainage in McMurray reservoir. J. Can.
Pet. Technol. 2006, 45, 54–62.
146. Zhao, D.W.; Wang, J.; Gates, I.D. Optimized solvent-aided steam-flooding strategy for recovery of thin heavy
oil reservoirs. Fuel 2013, 112, 50–59.
147. Yong, L.J.; Liu, M.J.; Hanqiao, Z.W.J. Optimizing injection parameters of polymer flooding in offshore
oilfields by combining surrogate model and genetic algorithm. China Offshore Oil Gas 2012, 4, 009.
148. Lei, Y.; Li, S.; Zhang, X.; Zhang, Q.; Guo, L. Optimal control of polymer flooding based on mixed-integer
iterative dynamic programming. Int. J. Control 2011, 84, 1903–1914.
149. Ekkawong, P. Multiobjective Design and Optimization of Polymer Flood Performance. Ph.D. Thesis, Texas
A&M University, College Station, TX, USA, 2013.
150. Al-Gosayir, M.; Babadagli, T.; Leung, J.Y.W. Optimization of Solvent Additive SAGD Applications
using Hybrid Optimization Techniques. In Proceedings of the SPE Enhanced Oil Recovery Conference,
Kuala Lumpur, Malaysia, 19–21 July 2011; Society of Petroleum Engineers: Kuala Lumpur, Malaysia, 2011.
151. Emera, M.K.; Sarma, H.K. Use of Genetic Algorithm to Predict Minimum Miscibility Pressure (MMP)
Between Flue Gases and Oil in Design of Flue Gas Injection Project. In Proceedings of the SPE Middle East
Oil and Gas Show and Conference, Kingdom of Bahrain, 12–15 March 2005; Kingdom of Bahrain Society of
Petroleum Engineers: Kingdom of Bahrain, 2005.
152. Dehghani, S.M.; Sefti, M.V.; Ameri, A.; Kaveh, N.S. Minimum miscibility pressure prediction based on a
hybrid neural genetic algorithm. Chem. Eng. Res. Des. 2008, 86, 173–185.
153. Chen, S.; Li, H.; Yang, D.; Tontiwachwuthikul, P. Optimal parametric design for water-alternating-gas (WAG)
process in a CO2 -miscible flooding reservoir. J. Can. Pet. Technol. 2010, 49, 75–82.
154. Sen, R. Biotechnology in petroleum recovery: The microbial EOR. Prog. Energy Combust. Sci. 2008, 34,
714–724.
155. Horowitz, B.; Guimarães, L.J.D.N.; Dantas, V.; Afonso, S.M.B. A concurrent efficient global optimization
algorithm applied to polymer injection strategies. J. Pet. Sci. Eng. 2010, 71, 195–204.
Processes 2017, 5, 34 25 of 25
156. Jones, D.R. A taxonomy of global optimization methods based on response surfaces. J. Glob. Optim. 2001,
21, 345–383.
157. Dixon, T.; Yamaji, K.; van t Veld, K.; Mason, C.F.; Leach, A. The Economics of CO2 Sequestration through
Enhanced Oil Recovery. Energy Proc. 2013, 37, 6909–6919.
158. Jahangiri, H.R.; Zhang, D. Ensemble based co-optimization of carbon dioxide sequestration and enhanced
oil recovery. Int. J. Greenh. Gas Control 2012, 8, 22–33.
159. Chen, Y.; Oliver, D.S. Ensemble-Based Closed-Loop Optimization Applied to Brugge Field. SPE Reserv.
Eval. Eng. 2010, 2013, 56–71.
160. Rwechungura, R.W.; Suwartadi, E.; Dadashpour, M.; Kleppe, J.; Foss, B.A. The Norne Field Case—A Unique
Comparative Case Study. SPE Intell. Energy Conf. Exhib. 2013, doi:10.2118/127538-MS.
161. Rwechungura, R.; Bhark, E.; Miljeteig, O.T.; Suman, A.; Kourounis, D.; Foss, B.; Hoierand, L.; Kleppe, J.
Results of the First Norne Field Case on History Matching and Recovery Optimization Using Production
and 4D Seismic Data. In Proceedings of the SPE Annual Technical Conference and Exhibition, San Antonio,
TX, USA, 8–10 October 2012; Society of Petroleum Engineers: San Antonio, TX, USA, 2012.
162. Chen, Y.; Oliver, D.S. History matching of the Norne full-field model with an iterative ensemble smoother.
SPE Reserv. Eval. Eng. 2014, 17, 244–256.
163. Suman, A.; Fernández-Martínez, J.L.; Mukerji, T. Joint Inversion of Time-Lapse Seismic and Production Data
for Norne Field. In Proceedings of the 2011 SEG Annual Meeting, San Antonio, TX, USA, 18–23 September
2011; Society of Exploration Geophysicists: San Antonio, TX, USA, 2011.
164. Hedengren, J.D.; Shishavan, R.A.; Powell, K.M.; Edgar, T.F. Nonlinear modeling, estimation and predictive
control in APMonitor. Comput. Chem. Eng. 2014, 70, 133–148.
165. Cao, F. Development of a Two-Phase Flow Coupled. Ph.D. Dissertation, University of Texas at Austin,
Austin, TX, USA, 2014.
c 2017 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).